Entropy and Clustering Information Applied to sEMG Classification.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Machine Learning

Background:

• Electromyographic (EMG) signals present stochastic complexity, making classification difficult.
• Reducing signal complexity is key for improving classification accuracy.
• Clustering techniques offer a method to simplify signal spaces for analysis.

Purpose of the Study:

• To investigate the effectiveness of clustering electromyographic signals for improved movement classification.
• To reduce the complexity of EMG signal classification by mapping signals to a smaller dimensional space.
• To enhance the performance of posterior classifiers by providing prior information through clustering.

Main Methods:

• Applied Agglomerative Hierarchical Clustering to electromyographic signals.
• Resized the signal into a smaller dimensional space using clustering.
• Compared clustered signals with possible human movements for classification.
• Utilized a short time window (10 ms) for signal analysis.

Main Results:

• Achieved classification accuracy exceeding 90%.
• Demonstrated significant improvement in classification by using clustering prior to the final classifier.
• The methodology proved competitive with existing literature methods.
• Effective classification was achieved with a signal sampled at 2000 Hz.

Conclusions:

• Clustering electromyographic signals effectively reduces classification complexity.
• Agglomerative Hierarchical Clustering provides a viable method for enhancing EMG signal classification accuracy.
• The proposed method offers a competitive and efficient approach for EMG-based movement recognition.